Predicting the Open-Hole Tensile Strength of Composite Plates Based on Probabilistic Neural Network

Tensile experiments were performed for open-hole composite plates with three different layups. With the limited number of experimental results, a probabilistic neural network (PNN) based approach is proposed to predict the tensile strength of composite plates with an open-hole. The predictive model takes the geometric parameters, the layup features and the average tensile stress of open-hole composite plates as the inputs and produces the safety status as the intermediate output with the classification function of PNN. Then the critical safety point, that is the open-hole tensile strength, where the safety status turns from survival to failure, is determined with the bi-section searching method. The predictions produce acceptable results whose errors are comparable to the coefficient of variation of experimental results. With experimental data from other studies, further assessments are also made to prove the capability of this model in predicting the open-hole tensile strength of composite plates.